No other price source publishes its measured error
How accurate are our prices, actually?
Every fair value on VaultStore is backtested against realized sales after the fact: did the price we published before the sale land close to what the card actually sold for? This page publishes that measurement directly — mean absolute percentage error (MAPE) and confidence-interval coverage — rather than asking you to trust a marketing claim. See the methodology page for how the underlying fair-value number is computed.
Measured accuracy (30d)
MAPE + CI coverage vs realized salesAccuracy metrics publish here as sample sizes reach the 50-paired-comp floor. No backtested sales have landed in this window yet. This is an honest “not enough data yet,” not a hidden number.
Freshness — how fast are we, really
% of catalog priced within 24h / 72hFreshness metrics publish here once priced catalog rows exist to measure against. No priced rows are available to score yet.
First-party coverage (90d)
How much of the catalog has a real VaultStore transaction behind itNo first-party comps (a real VaultStore order, auction settlement, or buylist fill) have landed in the trailing 90 days yet. Every displayed price today is still built from external vendor consensus rather than our own sales — the truthful starting point our market-making loop exists to change.
What this page will never do
- Publish a MAPE or coverage number below the 50-paired-comp floor.
- Report a lane’s new-set lag as measurable when the upstream source publishes no release date.
- Show operator-only internals — job health, environment flags, or bid-target backlogs stay on the admin surface.
- Fabricate a number when the underlying table is empty. An honest “not enough data yet” beats a fake one.